The biggest talk coming out of Selection Sunday was the selection of Iona as one of the last two at-large teams. This was generally not forseen, although their #41 RPI ranking would have gotten them selected had the selection committee gone straight by RPI. The issue is that, compared with Drexel (a team with a similar schedule difficulty and record), Iona had worse losses. The RPI does not take this into account, and consequently their RPI ranking was much higher than a probabilistic ranking that considers wins/losses and specific opponents (I rank them #61). On the flip side, Drexel was hurt by the simplification made in the RPI that opponents' record and opponents' opponents' record equals schedule strength. While their #66 RPI made it highly unlikely that they would be selected, an iterative RPI calculation ranks them #43.
Other surprise inclusions were West Virginia and South Florida, both of whom have RPIs in the 50s as well as rank #56 and #60 in my own rankings. Simply put, there is little statistical evidence to support either team's selection. South Florida, with 20 wins and the higher RPI (#53), was somewhat understandable. However, West Virginia didn't reach 20 wins and had the worst RPI of all selected teams (#56). Oral Roberts would have been a worthy replacement, with a better RPI than either team (#51) as well as the better ranking if specific games are considered (#46 in my ranking).
The worst-ranked team (in my system) selected was Colorado State, which took full advantage of both the schedule strength approximation (an iterative RPI would have ranked them #52 instead of #30), as well as the error implicit in ignoring the individual games (my ranking puts them at #66). This is not a knock on the selection committee, as excluding the #30 team in the RPI would have been nearly unprecedented.
Oddly enough, the team hurt the most by the RPI was not a midmajor as is typically expected. It was Oregon, whose RPI of #64 made them very unlikely to be selected. I rank them #47. Middle Tennessee State was also hurt, with an RPI of #57 but a ranking of #45 in my system.
Attempting to line up team RPI rankings with seedings, it seems that there was a uniform shift downwards by 1-2 seeds that was applied to nearly all mid-major teams. Wichita State (#12 RPI, #5 seed), Memphis (#15, #8), Creighton (#23, #8), Long Beach State (#34, #12), Harvard (#35, #12), and VCU (#39, #12) all were bumped below where they would have gone by the RPI. And, the only teams that were seeded significantly higher than the RPI indicated were major-conference teams (Wisconsin, Cincinnati, Notre Dame, Kansas State, Virginia, and West Virginia). If the NCAA believes the RPI to be systematically biased against major conferences, they should fix it. However, systematically giving better seeds to major conferences and worse ones to midmajors gives the impression of favoritism.
Note: if you use any of the facts, equations, or mathematical principles introduced here, you must give me credit.